In stock · Fully configured · USA
NVIDIA AI Servers Ready to Ship
Fully configured 8-GPU NVIDIA B300, B200 and H200 servers from ASRock Rack, ASUS, Supermicro, Lenovo and Dell. Add to cart today and ship now, while supplies last.
Ready-to-Ship NVIDIA GPU Servers for AI
Skip the months-long allocation queue. AVADirect has ready-to-ship NVIDIA B200 and NVIDIA H200 servers in stock in the USA, and NVIDIA B300 servers arriving at the end of November and in December 2026. Each one is fully configured with eight GPUs, high-core-count CPUs, DDR5 memory and fast NVMe storage.
NVIDIA B300 (Blackwell Ultra) systems deliver 288GB of HBM3e per GPU for the largest language models and long-context inference. NVIDIA B200 servers bring Blackwell performance to AI training and generative AI, and NVIDIA H200 servers offer proven Hopper performance for production inference and HPC. Inventory is limited and available while supplies last.
Supermicro AS-A126GS-TNBR
8× NVIDIA B200 180GB GPU Server
The Supermicro AS-A126GS-TNBR, fully configured with 8× NVIDIA B200 180GB GPUs, is a high-performance AI server designed for demanding AI training, deep learning, and high-performance computing workloads. Powered by dual AMD EPYC 9575F processors and equipped with 2,304GB of DDR5-6400 memory and high-speed NVMe storage, this system delivers exceptional computing capabilities for enterprise AI infrastructure, generative AI, and large language models. Fully configured and ready to ship, it offers a fast path to deploying next-generation NVIDIA Blackwell GPU computing.
Specifications
- Manufacturer
- Supermicro
- Model
- AS-A126GS-TNBR
- Form Factor
- 10U Rackmount
- GPU Platform
- NVIDIA HGX B200
- GPUs
- 8× NVIDIA B200
- GPU Memory
- 180GB HBM3e per GPU
- Total GPU Memory
- 1,440GB (1.44TB)
- Processors
- 2× AMD EPYC 9575F
- CPU Cores
- 64 cores per processor (128 cores total)
- System Memory
- 2,304GB DDR5-6400 ECC RDIMM
- Boot Storage
- 2× 960GB Phison D100P M.2 NVMe SSDs
- Data Storage
- 4× 3.84TB Samsung PM9A3 U.2 NVMe SSDs
- Total Raw Storage
- 17.28TB
- GPU Interconnect
- NVIDIA NVLink with NVSwitch
- Power Supplies
- 6× 5,250W redundant power supplies
Full product details SKU 20591655
Lenovo ThinkSystem SR680a V3
8× NVIDIA B200 180GB GPU Server
The Lenovo ThinkSystem SR680a V3 is a high-performance 8U AI server equipped with eight NVIDIA B200 180GB GPUs, delivering exceptional computing power for large-scale AI training, deep learning, and high-performance computing workloads. Powered by dual Intel Xeon 8558 processors and featuring 2TB of DDR5-5600 memory, this system combines advanced GPU acceleration with high-speed NVMe storage to support demanding enterprise AI applications, generative AI, and large language models. Fully configured and ready to ship, it provides a fast path to deploying next-generation NVIDIA Blackwell AI infrastructure.
Specifications
- Manufacturer
- Lenovo
- Model
- ThinkSystem SR680a V3
- Form Factor
- 8U Rackmount
- GPU Platform
- NVIDIA HGX B200
- GPUs
- 8× NVIDIA B200
- GPU Memory
- 180GB HBM3e per GPU
- Total GPU Memory
- 1,440GB (1.44TB)
- Processors
- 2× Intel Xeon 8558
- System Memory
- 2TB DDR5-5600
- Boot Storage
- 2× 960GB M.2 SSDs (RAID 1)
- Usable Boot Capacity
- 960GB
- Data Storage
- 4× 3.84TB U.3 NVMe SSDs
- Total Raw Data Storage
- 15.36TB
- Total Raw SSD Capacity
- 17.28TB
Full product details SKU 20591709
Dell PowerEdge XE9680
8× NVIDIA H200 141GB GPU Server
The Dell PowerEdge XE9680, fully configured with 8x NVIDIA H200 GPUs, delivers exceptional computing performance for demanding AI, machine learning, deep learning, and high-performance computing workloads. Designed for enterprise-scale AI infrastructure, this powerful GPU server provides the performance and scalability needed for complex workloads. Fully configured and ready to ship, it offers a fast path to deploying advanced AI computing infrastructure.
Specifications
- Server Model
- Dell PowerEdge XE9680
- GPUs
- 8× NVIDIA H200
- GPU Memory
- 141GB per GPU (1,128GB total)
- Processors
- 2× Intel Xeon 8580
- System Memory
- 2TB DDR5-5600
- Boot Storage
- 2× 960GB SSDs (RAID 1)
- Additional Storage
- 2× 3.84TB U.2 NVMe SSDs
- Total Raw SSD Capacity
- 9.6TB
Full product details SKU 20591435
ASRock Rack 8U16X-TURIN2 B300
8× NVIDIA B300 288GB GPU Server
The ASRock Rack 8U16X-TURIN2 B300 is a next-generation AI server featuring eight NVIDIA B300 288GB GPUs, delivering massive GPU memory capacity and computing performance for demanding AI training, inference, and high-performance computing workloads. Powered by dual AMD EPYC 9575F processors and equipped with 3,072GB of DDR5-6400 memory, this system combines advanced GPU acceleration with high-capacity NVMe storage and high-speed networking, including 8× NVIDIA ConnectX-8 adapters and a dual-port 200GbE ConnectX-7 adapter. Designed for enterprise AI infrastructure, large language models, and generative AI applications, this powerful system is fully configured, with an expected ship date of end of November 2026.
Specifications
- Manufacturer
- ASRock Rack
- Model
- 8U16X-TURIN2 B300
- Form Factor
- 8U Rackmount
- GPU Platform
- NVIDIA B300
- GPUs
- 8× NVIDIA B300
- GPU Memory
- 288GB per GPU
- Total GPU Memory
- 2,304GB (2.304TB)
- Processors
- 2× AMD EPYC 9575F
- System Memory
- 3,072GB DDR5-6400
- Boot Storage
- 2× 960GB M.2 SSDs
- Data Storage
- 4× 7.68TB U.2 NVMe SSDs
- Total Raw Data Storage
- 30.72TB
- Total Raw SSD Capacity
- 32.64TB
- High-Speed Networking
- 8× NVIDIA ConnectX-8
- Additional Networking
- 1× Dual-Port 200GbE NVIDIA ConnectX-7
Full product details SKU 20591767
ASUS XA NB3I-E12
8× NVIDIA B300 288GB GPU Server
The ASUS XA NB3I-E12 is a high-performance AI server designed for demanding enterprise AI training, inference, and high-performance computing workloads. Equipped with eight NVIDIA B300 288GB GPUs and powered by dual Intel Xeon 6767P processors, this system delivers exceptional GPU acceleration for generative AI, deep learning, and large language models. It features 128GB of DDR5-6400 memory in the supplied configuration, with an optional upgrade to 3,072GB, along with a 7.68TB U.2 PCIe 5.0 NVMe SSD for high-speed data access. Fully configured for advanced AI computing, it provides a powerful foundation for next-generation enterprise infrastructure. Expected to ship in December 2026.
Specifications
- Manufacturer
- ASUS
- Model
- XA NB3I-E12
- GPU Platform
- NVIDIA B300
- GPUs
- 8× NVIDIA B300
- GPU Memory
- 288GB per GPU
- Total GPU Memory
- 2,304GB (2.304TB)
- Processors
- 2× Intel Xeon 6767P
- System Memory
- 128GB DDR5-6400
- Optional Memory Upgrade
- 3,072GB DDR5-6400
- Storage
- 1× 7.68TB U.2 PCIe 5.0 NVMe SSD
- Storage Interface
- PCIe 5.0 NVMe
Full product details SKU 20591809
Compare NVIDIA B300, B200 & H200 Servers
| Server | GPUs | GPU memory | Processors | System memory | Storage | Networking | Form factor |
|---|---|---|---|---|---|---|---|
| Supermicro AS-A126GS-TNBR | 8× NVIDIA B200 | 180GB (1.44TB total) | 2× AMD EPYC 9575F | 2,304GB DDR5-6400 | 2× 960GB M.2 + 4× 3.84TB U.2 NVMe | — | 10U |
| Lenovo ThinkSystem SR680a V3 | 8× NVIDIA B200 | 180GB (1.44TB total) | 2× Intel Xeon 8558 | 2TB DDR5-5600 | 2× 960GB M.2 (RAID 1) + 4× 3.84TB U.3 NVMe | — | 8U |
| Dell PowerEdge XE9680 | 8× NVIDIA H200 | 141GB (1.13TB total) | 2× Intel Xeon 8580 | 2TB DDR5-5600 | 2× 960GB (RAID 1) + 2× 3.84TB U.2 NVMe | — | 6U |
| ASRock Rack 8U16X-TURIN2 B300 | 8× NVIDIA B300 | 288GB (2.3TB total) | 2× AMD EPYC 9575F | 3,072GB DDR5-6400 | 2× 960GB M.2 + 4× 7.68TB U.2 NVMe | 8× ConnectX-8 + 1× 200GbE ConnectX-7 | 8U |
| ASUS XA NB3I-E12 | 8× NVIDIA B300 | 288GB (2.3TB total) | 2× Intel Xeon 6767P | 128GB DDR5-6400 (3,072GB optional) | 1× 7.68TB U.2 PCIe 5.0 NVMe | — | 9U |
GPU Server FAQ
Which NVIDIA GPU servers are ready to ship?
The NVIDIA B200 and NVIDIA H200 servers are fully configured and ready to ship. The NVIDIA B300 servers are arriving soon: the ASRock Rack 8U16X-TURIN2 is expected at the end of November 2026 and the ASUS XA NB3I-E12 in December 2026, and you can order them now. Quantities are limited and orders are fulfilled while supplies last.
What is the difference between NVIDIA B300, B200 and H200?
NVIDIA B300 (Blackwell Ultra) has 288GB of HBM3e memory per GPU, or 2.3TB in an 8-GPU server. NVIDIA B200 (Blackwell) has 180GB per GPU, or 1.44TB per server. NVIDIA H200 (Hopper) has 141GB per GPU, or 1.13TB per server. More GPU memory lets you train and serve larger models with longer context windows.
Can I order more than one server?
Yes. Add a server to your cart and change the quantity in the cart, or contact our sales team for volume pricing and multi-node deployments.
Do you offer financing for GPU servers?
Yes. AVADirect offers business financing options. See our financing page or ask our sales team.
Questions? Talk to Our GPU Server Team
Need several nodes, a custom configuration, rack integration or financing? Send us a message and our enterprise sales team will get back to you, usually within one business day.
Prefer to talk? Call 1-866-766-4629
Prices are per server in USD, exclude tax and shipping, and are subject to availability and change without notice. Quantities are limited and available while supplies last. NVIDIA, HGX, Blackwell, Hopper and ConnectX are trademarks of NVIDIA Corporation. AMD and EPYC are trademarks of Advanced Micro Devices, Inc. Intel and Xeon are trademarks of Intel Corporation. Other product and company names are trademarks of their respective owners.